Skip to content

Which companies just had a verified incident, who lends to them and who owns them. Every claim carries its evidence level and its source. How we decide what is confirmed

DFX Cyber DeskDevelopingcloud PaaS / developer infrastructure · United Kingdom

Northflank API flaw let attacker decrypt customer secrets; stolen AI keys abused

On 2026-10-30 an attacker exploited an unknown vulnerability in cloud platform Northflank's API, decrypted stored secrets and exfiltrated those matching AI, cloud and crypto credential patterns. Northflank says 1.5% of customer teams were affected and customers reported misuse of AI provider keys.

Published · 3 min read · 3 sources

What happened

Northflank's incident report, published on 2026-11-17, says an attacker exploited a previously unknown vulnerability in Northflank's API to achieve remote code execution in one API component that had access to the database and the keys used to decrypt stored secrets. The attacker probed the platform from 2026-10-29, signed up at 00:19 UTC on 2026-10-30, achieved code execution at 02:39 and, between 02:41 and 03:08, read encryption keys and source code, queried the database, decrypted secrets and uploaded matching records. The attacker removed its tools by 08:12.2

According to Northflank, the attacker read the component's source code with standard command-line tools, appears to have used AI code generation to modify scripts in real time, and had a working decryption script within six minutes. The scripts uploaded only records matching credential patterns for AI providers, AWS, private keys, Slack, GitHub and GitLab tokens, and crypto wallet keys or seed phrases, or names suggesting such credentials.2

Who is affected

Northflank says 1.5% of customer teams had secret values extracted and 1.8% of secret values stored on the platform were exfiltrated. Affected features were project secret groups (almost all affected customers), global team secrets, notification integrations such as Slack, Discord and Microsoft Teams, and bring-your-own-cloud integrations that used static credentials. It has not published the number of customers.2

Northflank says it found no evidence that user credentials or customer personal data were exfiltrated, or that services, jobs, addons, builds, customer workloads or workload data were accessed. On its status page it told customers that forward-deployed or dedicated control planes were unaffected, while earlier advising all customers to assume every environment variable had been decrypted.1,2

On 2026-10-30 Northflank relayed that multiple users reported their AI API keys stored on the platform had been used without authorization. Its report says customer reports of misuse to date relate to AI provider keys. No customer has been named.1,2

Who is behind it

Northflank has not named an actor. It published indicators including an AWS EC2 address used for sign-up, command and control and exfiltration, and two sign-up email addresses on one domain. Wiz lists Northflank among more than a dozen AI developer platforms that disclosed incidents in October and November 2026 in a pattern it describes as targeted exploitation hunting stored environment variables, especially those granting AI access; Wiz does not attribute Northflank's incident to an actor.2,3

Financial and operational impact

No loss amount has been published, either by Northflank or for the abused AI keys. Northflank says the platform stayed operational and running workloads were not affected. Its status page showed the incident as a major outage from 2026-10-30 until it was marked resolved on 2026-11-17.1,2

Customers carried the remediation burden: Northflank told every customer to rotate AI, cloud, payment and database credentials and to check AI provider usage and billing from 2026-10-30. It revoked all Northflank API tokens, so customers who stored them as secrets had to create new ones.1,2

Response and recovery

Northflank's report says its security monitoring alerted the team at 11:43 UTC (the last attacker activity other than tool removal was at 03:11), and that the vulnerability was fixed at 13:05 UTC, 82 minutes after the alert. Its earlier status update had described the alert as coming from Wiz and the exploit as patched within 60 minutes of notification. The status page also said a separate vulnerability, CVE-2026-98432, was patched; the incident report does not mention that CVE or name the product behind it.1,2

Northflank rotated its own platform, database and object storage credentials, revoked cross-account links, API and VCS keys, and is re-encrypting stored secrets under new keys. It released self-service credential rotation for database, cache, object storage and message queue addons, made audit logs available to all customers, and engaged FTI Consulting on 2026-10-31. It says the investigation continues with Wiz's incident response team and FTI Consulting.1,2

What investors and lenders should watch

The sources read do not say who owns or funds Northflank. For customers and their backers, the exposure is credentials: any company that stored AI, cloud, payment or crypto keys on Northflank should be able to show it rotated them and reviewed provider bills after 2026-10-30. Watch for downstream disclosures from customers whose keys were used, for the forensic findings, and for whether the platform incidents Wiz grouped together, including Railway, are tied to a common actor.1,2,3

What is not known

  • The specific API vulnerability has not been described, and the product behind CVE-2026-98432 has not been named.
  • The absolute number of affected customers has not been disclosed, only a percentage of teams.
  • The cost and providers of the AI key abuse have not been disclosed.
  • No actor has been named.
  • The status page and the incident report differ on who raised the 11:43 UTC alert and on patch timing (60 versus 82 minutes).

Timeline

  1. Attacker first probes the platform, per Northflank.2

  2. 02:39 UTC remote code execution; 02:41 to 03:08 secrets decrypted and exfiltrated; 11:43 alert; 13:05 vulnerability fixed; 13:40 status page post.1,2

  3. Northflank relays customer reports of unauthorized use of AI API keys.1

  4. FTI Consulting engaged; audit logs and self-service secret rotation for addons released.1,2

  5. Status update: no evidence customer workloads or data were accessed; forensic review ongoing.1

  6. Incident report published; status page incident marked resolved.1,2

Video and further coverage

Reading video and press coverage...

Sources

  1. 1.Northflank status: Security incident (notice history)Northflank, 2026-11-17 · primary
  2. 2.Incident report: October 30 Security IncidentNorthflank, 2026-11-17 · primary
  3. 3.Black October for AI Developer PlatformsWiz, 2026-10-30 · research
Cite this report

DFX Cyber Desk, "Northflank API flaw let attacker decrypt customer secrets; stolen AI keys abused," DFX Intelligence, published 2026-11-19, https://dfxintel.com/cyber/reports/northflank

DFX Cyber Desk reports state only what the cited sources say. A ransomware group's claim is reported as a claim, losses are reported only as a source states them, and nothing is attributed without a named source. Methodology